Ultralytics YOLO

license

AGPL-3.0(OSI)

source

public(github.com/ultralytics/ultralytics)

core features withheld

yes — enterprise-only proprietary models supplied on request

The library and the public YOLO checkpoints are AGPL-3.0, and the whole package builds from the public repository. Ultralytics' paid Enterprise License is mostly a way out of the AGPL's share-alike terms, but its own terms also cover proprietary models trained on an Ultralytics enterprise dataset and supplied only to enterprise customers, so part of what Ultralytics ships is kept out of the open release.

• https://raw.githubusercontent.com/ultralytics/ultralytics/HEAD/LICENSE recorded 2026-09-26

  The LICENSE file is the full "GNU AFFERO GENERAL PUBLIC LICENSE Version 3, 19 November 2007" text.

• https://raw.githubusercontent.com/ultralytics/ultralytics/HEAD/README.md recorded 2026-09-26

  README License section: "Ultralytics offers two licensing options": AGPL-3.0, and an "Ultralytics Enterprise License" for development and production use "bypassing the open-source requirements of AGPL-3.0".

• https://ungh.cc/repos/ultralytics/ultralytics/files/main recorded 2026-09-26

  Repository tree at main: 1,036 files under .github, docker, docs, examples, tests and ultralytics, with no ee, enterprise, commercial or pro directory.

• https://www.ultralytics.com/legal/enterprise-software-license recorded 2026-09-26

  Enterprise Software License Terms: "Software also includes: - Proprietary Models: Access to the Company's proprietary models, available exclusively to enterprise customers and directly supplied by the Company."

• https://www.ultralytics.com/license recorded 2026-09-26

  Licensing page comparison row: "Access to enterprise models - Full access to pretrained, large-scale models trained on 1M-image Ultralytics Enterprise Dataset."

Measured on monthly PyPI downloads of the ultralytics package, which is how the library and its checkpoints are installed.

• https://pypistats.org/api/packages/ultralytics/recent recorded 2026-09-26

  last_month 4,028,814 downloads of ultralytics

Ultralytics covers most image-perception tasks from one training and inference interface, with pretrained checkpoints for each. Its fixed-class YOLO models need a training run for new object classes, but the same package runs the open-vocabulary YOLO-World and YOLOE detectors and SAM 3, which detect or segment what the user names in their own images with no training. It does not search for or train a model automatically.

• https://docs.ultralytics.com/ recorded 2026-09-26

  Docs landing: "YOLO is a family of real-time computer vision models for object detection, instance segmentation, semantic segmentation, depth estimation, classification, pose estimation, oriented bounding boxes, and tracking, available through one Python package and CLI."

• https://docs.ultralytics.com/tasks/ recorded 2026-09-26

  Tasks page sections: Detection, Instance Segmentation, Semantic Segmentation, Depth Estimation, Classification, Pose Estimation, OBB.

• https://ungh.cc/repos/ultralytics/ultralytics/files/main recorded 2026-09-26

  Repository tree at main includes ultralytics/models/yolo/world/, ultralytics/models/yolo/yoloe/ (with predict.py) and ultralytics/models/sam/, and the docs pages models/yolo-world.md, models/yoloe.md and models/sam-3.md.